stp2y - CybAI news

24 May

Meta and Google want to make AI deals with Hollywood studios

stp2y0 CommentsNewsgear, google, meta, news, openai

Meta and Google are offering Hollywood studios millions of dollars with the hope of striking licensing deals that could improve their models for AI-generated video, according to a in Bloomberg. The companies have reportedly offered “tens of millions of dollars,” though it’s unclear what will come from the talks.According to the report, Netflix and Disney “aren’t willing to license their content” but have “expressed interest in other types of collaborations.” Warner Brothers Discovery has reportedly indicated “a willingness to license some of its programs.”A spokesperson for Meta declined to comment. Google didn’t immediately respond to a request for comment. The…

24 May

The rise of intelligent automation as a strategic differentiator – AI News

stp2y0 CommentsChat-GPT

Intelligent automation (IA) technologies are graduating from being operational to highly strategic. In terms of the bottom line, it’s even more impressive. A study from SS&C Blue Prism, conducted by Forrester Consulting and published in April, put together a composite organisation representative of five customers interviewed. The conclusion was that, over three years, there were key gains in IA from greater productivity to compliance cost avoidance, to improved employee experience and retention. This represented an overall net present value of $53.4 million (£42.5m) per customer. Yet this may just be the tip of the iceberg. Dan Segura, enterprise sales manager…

24 May

Introducing Deephaven’s new calendar API | Deephaven

stp2y0 CommentsNews

Deephaven 0.32.0 introduced a number of new features, as well as improvements to old ones. Key features of the overhauled calendar API include:An updated Java calendar API that boosts performance in table operations.A streamlined Python API designed to minimize the chances of calling Python from query strings.Simple and intuitive naming conventions for ease of use.Support for user-defined calendars.Conversion between Deephaven and NumPy calendar formats.The new Python API is designed to be minimal, and facilitate the use of the Java API in query strings. The new calendar class provides the following methods:MethodDescriptionadd_calendarAdds a new business calendar to the set of available…

24 May

Goal Representations for Instruction Following

stp2y0 CommentsAIinstruction following, manipulation, representation learning

Goal Representations for Instruction Following A longstanding goal of the field of robot learning has been to create generalist agents that can perform tasks for humans. Natural language has the potential to be an easy-to-use interface for humans to specify arbitrary tasks, but it is difficult to train robots to follow language instructions. Approaches like language-conditioned behavioral cloning (LCBC) train policies to directly imitate expert actions conditioned on language, but require humans to annotate all training trajectories and generalize poorly across scenes and behaviors. Meanwhile, recent goal-conditioned approaches perform much better at general manipulation tasks, but do not enable easy…

24 May

Leaked OpenAI Documents Show Sam Altman Was Clearly Aware of Silencing Former Employees

stp2y0 CommentsChat-GPT

OpenAI's credibility — and the credibility of its CEO, Sam Altman — is crumbling.Last week, amid a surprise string of high-profile executive and safety team departures, Vox revealed that the ChatGPT creator had pressured employees into signing draconian non-disclosure and non-disparagement agreements by threatening to claw back exiting OpenAI employees' vested equity in the multibillion-dollar AI company.Clawing back vested equity — in short, the amount of company ownership that an employee has gained through their months or years of working there — is a highly unusual practice to begin with. This is especially true in the startup-powered Silicon Valley, where…

24 May

VINC-S: Closed-form Optionally-supervised Knowledge Elicitation with Paraphrase Invariance

stp2y0 CommentsGenAI

$^*$Equal contribution In Spring 2023, a team at EleutherAI and elsewhere worked on a follow-up to CCS that aimed to improve its robustness, among other goals. We think the empirical side of the project was largely unsuccessful, failing to provide evidence that any method had predictably better generalization properties. In the spirit of transparency, we are sharing our proposed method and some results on the Quirky Models benchmark. Introduction# As we rely more and more on large language models (LLMs) to automate cognitive labor, it's increasingly important that we can trust them to be truthful. Unfortunately, LLMs often reproduce human…

24 May

Judge Hints at Plans to Rein In Google’s Illegal Play Store Monopoly

stp2y0 CommentsRAG modelsandroid, antitrust, apps, courts, epic games, google, mobile, tech policy and law

A jury in December found that Google broke US antitrust laws through deals and billing rules that gave an unfair boost to its Google Play app store. On Thursday, a judge began laying out how Google could be forced to change its business as a penalty. The remedies under consideration could drive the most consequential shakeup ever to Google’s dominance over the Android universe.Fortnite video game developer Epic Games, which beat Google in the trial that saw a jury declare the Play store an illegal monopoly, is demanding that federal judge James Donato ban Google from contracts that deter competition.…

24 May

Text Generation using GPT2

stp2y0 CommentsViral News

The first ever GPT model was released by OpenAI in 2018. Since then, we have seen tremendous research and models based on the same architecture. Although GPT1 is old by today’s standards, GPT2 still holds fairly well for many fine-tuning tasks. In this article, we will dive into fine-tuning GPT2 for text generation. In particular, we will teach the model to generate detective stories based on Arthur Conan Doyle’s Sherlock Holmes series. Figure 1. Text generation example using the trained GPT2 model. GPTs and similar large language models can be fine-tuned for various text generation tasks. With this article, we…

24 May

4DS Plots ReRAM Roadmap – EE Times

stp2y0 CommentsNews

//php echo do_shortcode('[responsivevoice_button voice="US English Male" buttontext="Listen to Post"]') ?> 4DS Memory Limited has broken its radio silence to lay out its go-forward plans for its Resistive RAM (ReRAM) technology. The company said its interface switching capabilities based on PCMO (Praseodymium, Calcium, Manganese, Oxygen) delivers significant advantages over other filamentary ReRAM technologies, making its high-bandwidth, high-endurance persistent memory suitable for AI, big data and neural net applications. In a briefing with EE Times, chief strategy officer Peter Himes said 4DS’ ReRAM requires no refresh within its persistence window and can be “refreshed” within the DRAM operating window, which makes it…

24 May

Efficient and cost-effective multi-tenant LoRA serving with Amazon SageMaker | Amazon Web Services

stp2y0 CommentsAI

In the rapidly evolving landscape of artificial intelligence (AI), the rise of generative AI models has ushered in a new era of personalized and intelligent experiences. Organizations are increasingly using the power of these language models to drive innovation and enhance their services, from natural language processing to content generation and beyond. Using generative AI models in the enterprise environment, however, requires taming their intrinsic power and enhancing their skills to address specific customer needs. In cases where an out-of-the-box model is missing knowledge of domain- or organization-specific terminologies, a custom fine-tuned model, also called a domain-specific large language model…